RESUMO
Artificial intelligence (AI) systems for computer-aided diagnosis and image-based screening are being adopted worldwide by medical institutions. In such a context, generating fair and unbiased classifiers becomes of paramount importance. The research community of medical image computing is making great efforts in developing more accurate algorithms to assist medical doctors in the difficult task of disease diagnosis. However, little attention is paid to the way databases are collected and how this may influence the performance of AI systems. Our study sheds light on the importance of gender balance in medical imaging datasets used to train AI systems for computer-assisted diagnosis. We provide empirical evidence supported by a large-scale study, based on three deep neural network architectures and two well-known publicly available X-ray image datasets used to diagnose various thoracic diseases under different gender imbalance conditions. We found a consistent decrease in performance for underrepresented genders when a minimum balance is not fulfilled. This raises the alarm for national agencies in charge of regulating and approving computer-assisted diagnosis systems, which should include explicit gender balance and diversity recommendations. We also establish an open problem for the academic medical image computing community which needs to be addressed by novel algorithms endowed with robustness to gender imbalance.
Assuntos
Conjuntos de Dados como Assunto/normas , Aprendizado Profundo/normas , Interpretação de Imagem Radiográfica Assistida por Computador/normas , Radiografia Torácica/normas , Viés , Feminino , Humanos , Masculino , Padrões de Referência , Fatores SexuaisRESUMO
OBJECTIVE: To determine the interrater reliability (IRR) of lung ultrasonography (LUS) and chest radiography (CXR) and evaluate the accuracy of LUS compared with CXR for detecting pediatric pneumonia compared with chest computed tomography (CT) scan. STUDY DESIGN: This was a prospective cohort study of children aged 3 months to 18 years with a CXR and LUS performed between May 1, 2012, and January 31, 2014 with or without a clinical diagnosis of pneumonia. Four pediatric radiologists blinded to clinical information reported findings for the CXR and LUS images. IRR was estimated for 50 LUS and CXR images. The main outcome was the finding from CT ordered clinically or the probability of the CT finding for patients clinically requiring CT. Two radiologists reviewed CT scans to determine an overall finding. Latent class analysis was used to evaluate the sensitivity and specificity for findings (eg, consolidation) for LUS and CXR compared with CT. RESULTS: Of the 132 patients in the cohort, 36 (27%) had CT performed for a clinical reason. Pneumonia was clinically documented in 47 patients (36%). The IRR for lung consolidation was 0.55 (95% CI, 0.40-0.70) for LUS and 0.36 (95% CI, 0.21-0.51) for CXR. The sensitivity for detecting consolidation, interstitial disease, and pleural effusion was statistically similar for LUS and CXR compared with CT; however, specificity was higher for CXR. The negative predictive value was similar for CXR and LUS. CONCLUSIONS: LUS has a sufficiently high IRR for detection of consolidation. Compared with CT, LUS and CXR have similar sensitivity, but CXR is more specific for findings indicating pneumonia.
Assuntos
Pneumonia/diagnóstico por imagem , Radiografia Torácica/normas , Ultrassonografia/normas , Adolescente , Criança , Pré-Escolar , Feminino , Humanos , Lactente , Masculino , Variações Dependentes do Observador , Estudos Prospectivos , Reprodutibilidade dos Testes , Tomografia Computadorizada por Raios XRESUMO
OBJECTIVE: The purpose of this study was to compare the diagnostic accuracy achieved with and without the calibration method established by the DICOM standard in both medical-grade gray-scale displays and consumer-grade color displays. MATERIALS AND METHODS: This study involved 76 cases, six radiologists, three displays, and two display calibrations for a total of 2736 observations in a multireader-multicase factorial design. The evaluated conditions were interstitial opacities, pneumothorax, and nodules. CT was adopted as the reference standard. One medical-grade gray-scale display and two consumer-grade color displays were evaluated. Analyses of ROC curves, diagnostic accuracy (measured as AUC), accuracy of condition classification, and false-positive and false-negative rate comparisons were performed. The degree of agreement between readers was also evaluated. RESULTS: No significant differences in image quality perception by the readers in the presence or absence of calibration were observed. Similar forms of the ROC curves were observed. No significant differences were detected in the observed variables (diagnostic accuracy, accuracy of condition classification, false-positive rates, false-negative rates, and image-quality perception). Strong agreement between readers was also determined for each display with and without calibration. CONCLUSION: For the chest conditions and selected observers included in this study, no significant differences were observed between the three evaluated displays with respect to accuracy performance with and without calibration.
Assuntos
Terminais de Computador/estatística & dados numéricos , Terminais de Computador/normas , Apresentação de Dados/normas , Radiografia Torácica/instrumentação , Radiografia Torácica/normas , Software/normas , Colômbia , Cor , Desenho de Equipamento , Segurança de Equipamentos , Humanos , Variações Dependentes do Observador , Guias de Prática Clínica como Assunto , Radiografia Torácica/estatística & dados numéricos , Valores de Referência , Reprodutibilidade dos Testes , Sensibilidade e Especificidade , Software/estatística & dados numéricosRESUMO
OBJECTIVE: To assess interobserver agreement between thoracic radiologists, radiology residents, and emergency physicians in diagnosing pulmonary embolism (PE). MATERIALS AND METHODS: Emergency physicians, radiology residents, and thoracic radiologists evaluated 123 computed tomography pulmonary angiography images. Interobserver agreement was analysed using kappa statistics. RESULTS: Very good agreements were observed between thoracic radiologists and radiology residents (0.81 and 0.82). Fair and moderate agreements (0.39 and 0.42) were demonstrated between emergency physicians and thoracic radiologists. CONCLUSIONS: Important differences in interobserver agreement were found, with emergency physicians tending to overdiagnose PE.
Assuntos
Embolia Pulmonar , Radiologia , Humanos , Angiografia , Medicina de Emergência/normas , Internato e Residência , Pulmão/diagnóstico por imagem , Variações Dependentes do Observador , Embolia Pulmonar/diagnóstico por imagem , Interpretação de Imagem Radiográfica Assistida por Computador , Radiografia Torácica/normas , Radiologia/educação , Radiologia/normas , Estudos Retrospectivos , Tomografia Computadorizada por Raios XRESUMO
BACKGROUND: The vervet monkey (Chlorocebus sabaeus) is used commonly in cardiorespiratory biomedical research. This study was performed to establish reference values for thoracic structures and to describe the normal radiographic appearance of the vervet monkey thorax. METHODS: Right lateral and dorsoventral thoracic radiographs of ten mature vervet monkeys were evaluated. Anatomic structures were characterized using descriptive statistics. RESULTS: Normal measurements of skeletal, pulmonary, mediastinal, and cardiovascular structures are reported herein. Several ratios were calculated to assess the cardiac silhouette, caudal vena cava, and pulmonary arteries and veins. CONCLUSIONS: Consistent measurements could be made on the majority of the thoracic structures evaluated. The aorta on lateral radiographs and the pulmonary veins on dorsoventral radiographs were obscured by a mild bronchointerstitial pattern and body conformation. Caudal vena cava-tapering was occasionally noted and attributed to general anesthesia. Species-specific thoracic radiographic reference values should prove useful in vervet monkey disease diagnosis and management.
Assuntos
Chlorocebus aethiops/anatomia & histologia , Radiografia Torácica , Tórax/anatomia & histologia , Animais , Sistema Cardiovascular/anatomia & histologia , Feminino , Masculino , Interpretação de Imagem Radiográfica Assistida por Computador/normas , Radiografia Torácica/normas , Valores de Referência , Sistema Respiratório/anatomia & histologiaRESUMO
Se llevó a cabo un estudio descriptivo, longitudinal y prospectivo de 84 ancianos con insuficiencia cardíaca atendidos en el Servicio de Geriatría del Hospital Provincial Docente Dr Ambrosio Grillo Portuondo de Santiago de Cuba desde el 1 de julio del 2001 hasta igual fecha del 2002, para establecer el valor de los determinantes clínicos de mal pronóstico en estos pacientes. Entre los principales hallazgos se encontró que la aplicación de los criterios de Framinghan mejoró cualitativamente el diagnóstico. Hubo una vinculación estadísticamente significativa entre estos determinantes y la evaluación desfavorable de los pacientes.
A descriptive and prospective study was carried out in 84 aged people with heart failure attended at the Geriatrics Service of Dr Ambrosio Grillo Portuondo Provincial Teaching Hospital of Santiago de Cuba from July, 2001 to July, 2002, to establish the value of clinical determinants of poor prognosis in these patients. Among the main findings it was found that the application of the Framinghan criteria improved the diagnosis qualitatively. There was a statistically significant relation between these determinants and the unfavorable evaluation of patients.